Prediction of Acetone Levels in Gas Mixtures with High Ethanol Concentration: Investigating its Relevance for Diabetes Detection in Exhaled Air

Anna Paleczek, Dominik Grochala, Artur Rydosz

AGH University of Krakow

Poster PDF

Currently, the number of people with diabetes is growing rapidly, and early diagnosis and monitoring of the disease plays a key role in its treatment. Research on non-invasive determination of blood glucose levels on the basis of exhaled air is still underway. Previous studies do not take into account the influence of interfering factors on the response of gas sensors. One of them is ethanol, which shows a similar response mechanism of semiconductor gas sensors as acetone, which is a known biomarker of diabetes. As part of laboratory studies, we examined the effect of the presence of ethanol in gas mixtures that mimic human breath on the ability to predict the concentration of acetone in gas mixtures. The results showed that in mixtures without ethanol, the e-nose composed of gas sensors and the use of machine learning algorithms mean absolute error of predicting the acetone concentration was 0.25 ppm, and in the case of the presence of ethanol in the mixture, the error increased to 0.36 ppm. The conducted research emphasizes the importance of identifying influencing factors and examining their impact on the operation of e-nose-based systems. Recognizing and understanding these factors can improve the quality and reliability of breath analysis as a diagnostic tool. 

Affiliations: Institute of Electronics, Biomarkers Analysis LAB, AGH University of Science and Technology, Krakow, Poland